New data from Anthropic has revealed the extent to which AI usage is spreading unequally, both in the U.S. and across the world.
The company’s third Economic Index report tracked how and where its chatbot, Claude, is being utilized — and some clear trends have started to emerge.
Most notably, the report shows how location plays a major role in AI adoption.
Globally, the U.S. engages with Claude more than any other nation, claiming a 21.6% share of usage, well ahead of India in second place (7.2%), and Brazil, Japan and South Korea (all at 3.7%).
However, the results are not entirely unexpected, given these countries are large population centers. More insight is provided by adjusting each country’s share of usage by its share of the world’s working population — and in doing so, Anthropic uncovers a different picture.
Using these criteria, two wealthy countries with relatively small populations, Singapore and Israel, lead the way in their engagement with Claude, while other similar countries with strong economies such as Estonia, Malta, Luxembourg and Switzerland are well represented. In contrast, emerging economies such as India, Indonesia and Nigeria find themselves way behind in terms of usage per capita.
This can be explained by income — countries that use Claude most are likely to have high GDP per capita, robust internet connectivity and economies oriented more toward knowledge work than manufacturing.
But the company also sounds a note of warning, stating this disparity raises the issue of major economic divergence, as was the case with previous significant breakthroughs such as electrification or the combustion engine.
“If the effects of AI prove to be largest in richer countries, this general purpose technology might have similar economic implications,” the report stated.
The first-ever breakdown of population-adjusted usage of Claude in the U.S. reveals a notable geographical split, with northeastern states (New York, Massachusetts, Vermont, District of Columbia) and west coast states (California, Washington and Oregon) all in the top 25% and a string of southern ones (Oklahoma, Louisiana, Mississippi and Alabama) in the bottom 25%.
However, while there is also a clear link between usage and per capita GDP in the U.S., greater variance in the results than globally suggests other factors are at play. Anthropic believes this may be down to the nature of the work that powers individual states’ economies, highlighting that Washington D.C., which tops the usage index, has disproportionately frequent requests for editing documents and searches for information, while coding tasks are particularly common in California (third overall).
Among other areas explored in the data are the nature of tasks requested of Claude, the split between automation and augmentation, and which occupations are making the most use of AI. The data itself was compiled in August from around a million conversations.
Aside from the warning about uneven usage and its economic implications, Anthropic believes the report provides reason for optimism. “The nature of people’s use of Claude is evidently still being defined,” the report stated. “So far, though, it looks like we are becoming increasingly comfortable with AI.”